Ampere and Qualcomm aren’t the obvious of companions. Each, in any case, provide Arm-based chips for working knowledge middle servers (although Qualcomm’s largest market stays cell). However as the 2 firms introduced immediately, they’re now combining forces to supply an AI-focused server that makes use of Ampere’s CPUs and Qualcomm’s Cloud AI 100 Extremely AI inferencing chips for working — not coaching — fashions.
Like each different chip producer, Ampere is seeking to revenue from the AI increase. The corporate’s focus, nonetheless, has all the time been on quick and power-efficient server chips, so whereas it will possibly use the Arm IP so as to add a few of these options to its chips, it’s not essentially a core competency. That’s why Ampere determined to work with Qualcomm (and SuperMicro to combine the 2 options), Arm CTO Jeff Wittich tells me.
“The idea here is that while I’ll show you some great performance for Ampere CPUs running AI inferencing on just the CPUs, if you want to scale out to even bigger models — multi-100 billion parameter models, for instance — just like all the other workloads, AI isn’t one size fits all,” Wittich advised TechCrunch. “We’ve been working with Qualcomm on this solution, combining our super efficient Ampere CPUs to do a lot of the general purpose tasks that you’re running in conjunction with inferencing, and then using their really efficient cards, we’ve got a server-level solution.”
As for partnering with Qualcomm, Wittich stated that Ampere needed to place collectively best-of-breed options.
“[R]eally good collaboration that we’ve had with Qualcomm here,” he stated. “This is one of the things that we’ve been working on, I think we share a lot of really similar interests, which is why I think that this is really compelling. They’re building really, really efficient solutions and a lot of different parts of the market. We’re building really, really efficient solutions on the server CPU side.”
The Qualcomm partnership is a part of Ampere’s annual roadmap replace. A part of that roadmap is the brand new 256-core AmpereOne chip, constructed utilizing a contemporary 3nm course of. These new chips will not be fairly usually out there but, however Wittich says they’re prepared on the fab and may roll out later this yr.
On high of the extra cores, the defining function of this new technology of AmpereOne chips is the 12-channel DDR5 RAM, which permits Ampere’s knowledge middle prospects to higher tune their customers’ reminiscence entry in response to their wants.
The gross sales pitch right here isn’t simply efficiency, although, however the energy consumption and price to run these chips within the knowledge middle. That’s very true with regards to AI inferencing, the place Ampere likes to check its efficiency in opposition to Nvidia’s A10 GPUs.
It’s price noting that Ampere isn’t sunsetting any of its present chips in favor of those new ones. Wittich pressured that even these older chips nonetheless have loads of use circumstances.
Ampere additionally introduced one other partnership immediately. The corporate is working with NETINT to construct a joint resolution that pairs Ampere’s CPUs with NETINT’s video processing chips. This new server will be capable to transcode 360 reside video channels in parallel, all whereas additionally utilizing OpenAI’s Whisper speech-to-text mannequin to subtitle 40 streams.
“We started down this path six years ago because it is clear it is the right path,” Ampere CEO Renee James stated in immediately’s announcement. “Low power used to be synonymous with low performance. Ampere has proven that isn’t true. We have pioneered the efficiency frontier of computing and delivered performance beyond legacy CPUs in an efficient computing envelope.”
